采摘机器人果实识别与定位研究——基于双目视觉和机器学习

Research on Fruit Recognition and Location of Harvesting Robot Based on Binocular Vision and Machine Learning

分析了双目视觉系统的工作原理及视觉标定方法,利用YOLO V2卷积神经网络算法实现对目标果实的识别,并对目标果实的空间定位进行了深入研究,设计了一套基于双目视觉和机器学习的采摘机器人果实识别与定位系统。在多次实际定位实验中,橘子的深度定位误差最大值为1.06mm,证实了系统具有一定的准确性和稳定性。

It first analyzes the working principle of binocular vision system and vision calibration method, then realizes the recognition of the target fruit by using the algorithm of YOLO V2 convolution neural network, and makes a deep research on the spatial location of the target fruit, designs and studies a set of fruit recognition and location of picking robot based on binocular vision and machine learning. In many practical positioning experiments, the maximum depth positioning error of orange is 1.06 mm, which proves that the system has certain accuracy and stability.